[hami9293]

Hello everyone,

I am trying to solve a problem like
\nabla(D\nabla T) +f =0
where D is a function of x and y, (x,y being the coordinate system) with some well defined boundary conditions (Dirichlet or Neuman).

I want to have a simple FDM discretization like Gauss Seidel or SOR method, but since D is varying spatially I doubt that my simple scheme can work.
Does anyone know any better solution?

Thanks,

[FMDenaro]

Originally Posted by hami9293

Hello everyone,

I am trying to solve a problem like
\nabla(D\nabla T) +f =0
where D is a function of x and y, (x,y being the coordinate system) with some well defined boundary conditions (Dirichlet or Neuman).

I want to have a simple FDM discretization like Gauss Seidel or SOR method, but since D is varying spatially I doubt that my simple scheme can work.
Does anyone know any better solution?

Thanks,

I do not understand your question... GS, SOR are iterative algorithms for solving algebric linear systems, are not FDM discretization...

In second order discretization you can proceed for example:

[(D* dT/dx)|i+1/2-(D* dT/dx)|i-1/2]/dx